Support for alternative diffusivities

[jhdark]

FESTIM should be able to support different types of diffusivities and accept fem.Functions if given rather than just as floats or D_0 and E_D.

Guess this could link in with #742

[RemDelaporteMathurin]

Indeed, the Arrhenius form should be a special case (convenience). F.Material should be able to accept a diffusivity D as:

• a float
• a callable of space, time, temperature: eg. lambda x: 2 * x[0] or lambda x, T: D_0(x) * ufl.exp(-E_D/F.k_B/T)
• a dolfinx.fem.Function